From Pixels to Prognosis: A Deep Dive into Lung Cancer Subtype Classification using Transfer learning

Current diagnostic procedures, such as imaging tests and biopsies, are time intensive and prone to human error. As a result, we employed deep learning to uncover patterns and identify lung cancer from histology pictures. We hypothesised that combining two separate methods (transfer learning and data augmentation) would increase the efficacy of lung cancer detection. The study's dataset consisted of histology images for Lung Cancer and was quite unbalanced, with very few photos containing tumours. We hypothesised that transfer learning, which employs existing knowledge for classification and detection, would be effective on such an imbalanced dataset. To boost our learning, we employed data augmentation to incorporate changes into the dataset. This can be accomplished by implementing Sequential Keras Model Layers. We then tested the efficacy of various convolutional neural network models on this job. We achieved an accuracy of 87.50% utilising the transfer learning model MobileNetV2 as the basis model and numerous geometric changes, beating the cutting-edge convolutional neural network-based technique. We then analysed the results and published them on a website that people can access at any time and from any location.